supplementary information - research · 2018. 9. 11. · supplementary information the genomics of...
TRANSCRIPT
Supplementary Information
The genomics of selection in dogs and the parallel evolution between
dogs and humans
Guo-dong Wang*, Weiwei Zhai
*, He-chuan Yang, Ruo-xi Fan, Xue Cao, Li Zhong, Lu
Wang, Fei Liu, Hong Wu, Lu-guang Cheng, Andrei D. Poyarkov, Nikolai A. Poyarkov
JR, Shu-sheng Tang, Wen-ming Zhao, Yun Gao, Xue-mei Lv, David M. Irwin, Peter
Savolainen, Chung-I Wu¶, Ya-ping Zhang
¶
* These authors contributed equally to this work.
¶ Correspondence: [email protected] and [email protected]
This file includes:
Supplementary Figure S1 to S11
Supplementary Tables S1 to S7
Supplementary Note 1-5
Supplementary References
Supplementary Figures
Supplementary Figure S1: Data flow of our sequencing and analysis.
Supplementary Figure S2. False positives and false negatives in SNP calling. a) False
negative for the SNP calling when there are x individuals (x axis). Singleton and non-
singleton SNPs are also separately calculated. b) False positives for the SNP calling when
there are x individuals (x axis). c) Fitted linear relationship between the sample size and
the false negatives. d) Fitted linear relationship between the sample size and the false
positives.
Supplementary Figure S3. LDR distributions across the 11 individual genomes. The
genetic diversity across 38 autosomal and the sex chromosomes are plotted for each
individual. The low diversity regions (cutoff is set as 0.00005) are plotted in dark blue
regions.
Supplementary Figure S4. Structure analysis with K=2 and K=3. The cluster of 11
individuals using STRUCTURE by partitioning the sample into 2 or 3 groups,
respectively. Each vertical column represents an individual and each color represents a
genetic component.
Supplementary Figure S5. SNP ascertainment biases in the 48K SNP chip. a) The
site frequency spectra (polarized with an outgroup species) for these 48K SNPs in the
wolves. b) The site frequency spectra for these 48K SNPs in the Chinese native dogs. c)
The site frequency spectra for all the SNPs in the wolves. d) The site frequency spectra
for all the SNPs in the Chinese native dogs.
Supplementary Figure S6. Breeds that correlate with the PC2. This is the same plot
as Figure 2d, but highlighting the dog breeds that is further away from the rest of the
groups in the second principle component (PC2).
Supplementary Figure S7. PCA plot across 87 populations. The PCA plot for 87
populations using frequency spectra (including wolves and dogs sequenced in this study)
are plotted in a two dimensional plot for the first two principle components.
Supplementary Figure S8. The maximum likelihood estimates of divergence time
over the bootstrap samples. The upper bound of t is set to be 0.5. The x axis is the point
estimate for the divergence time (t), the y axis is the count (over 100 replicates).
Supplementary Figure S9. Isolation and migration model. a) the cartoon illustrating
of the IM(Isolation Migration) model we fitted to the data. b) the joint 2D site frequency
spectra observed in the data (heatmap). c) the predicted joint 2D site frequency spectra
from the best fitted model
Supplementary Figure S10. Site frequency spectra estimated using two different
methods. The site frequency spectra (SFS) inferred using two methods. The maximum
likelihood is extracted by using a program from Kim et al 2011. Top panels are for
wolves, bottom panels are from dogs.
Supplementary Figure S11. Fst and mean diversity across the dog and wolf genome. The left panels are the mean Fst across the genome. The right panels are the mean genetic
diversity for wolves (in red) and dogs (in blue). Window size is set to be 100kb and step
size is 20kb. Dashed lines are the cutoffs for picking the potentially positively selected
regions, which are also plotted with the purple horizontal bars.
Supplementary Tables
Supplementary Table S1. Library sizes and throughput for different individuals
Individual Location InsertSize Std Reads
Length
Matching
Bps
(Gbp)
GW1 Altai, Russia
409.38 124.03 111 5.51
340.76 39.47 120 16.91
217.64 91.49 111 0.09
339.82 43.87 81 4.29
GW2 Chukotka, Russia
333.82 25.76 120 18.79
297.69 129.2 111 1.31
286.46 105.46 111 1.18
GW3 Bryansk, Russia 329.62 77.74 101 26.48
GW4 Inner Mongolia,
China
2257.61 337.58 44 2.96
177.79 11.84 44 1.72
521.58 14.24 44 3.46
491.09 12.37 44 0.90
176.91 12 75 3.85
520.09 14.53 75 2.18
490.66 12.71 75 5.08
- - 44 2.77
dogCI1 Xi'an, China
371.95 36.35 101 15.31
421.37 51.38 101 12.38
485.13 101.3 101 4.67
dogCI2 Simao, China 313.04 19.3 171 23.44
dogCI3 Ya'an, China 332.45 21.67 121 24.44
dogTM Lijiang, China 314.36 82.51 101 24.73
dogGS Germany 478.97 17.17 44 5.71
373.11 116.52 65 17.10
dogBM France 488.61 16.17 44 5.32
472.56 22.47 90 18.78
Supplementary Table S2. The Sanger read coverage distribution across six
individuals
Coverage 1 2 3 4 5 6
region_length 4923 2243 2792 5688 8806 29866
Supplementary Table S3. False positive and false negative for SNP calling
Sample Sanger_snp GAIIx overlap false_positive false_negativea
GW3 206 170 166 0.024 0.194
GW4 251 200 199 0.005 0.207
dogCI1 211 180 176 0.022 0.166
dogGS 152 113 112 0.009 0.263
dogTM 194 158 155 0.019 0.201
dogBM 210 159 157 0.013 0.252
Sum 1224 980 965 0.015 0.212 aA large proportion of the false negatives are due to low coverage in the sequencing
Supplementary Table S4. False positive and false negative rate for indel calling
sanger_indels solexa_indels Overlap false_positive false_negative
222 107 100 0.065 0.55
Supplementary Table S5. Demographic inference with different upper bound
Parameter Estimated value nonparametric bootstrap
confidence intervala
nonparametric bootstrap
confidence intervalb
Nanc 53,000 - -
Nwolf 50,000 (49,543, 50,163) (49,522,50,190)
Ndog_bottleneck 8,500 (8,161, 8,961) (8,089, 8956)
Ndog_current 17,000 (16,552, 17,585) (16,573, 17,692)
T 32,000 (years) (31,300, 33,191) (31,158, 33,137)
Mwd 1.31 (1.24, 1.40) (1.22, 1.41)
Mdw 1.71 (1.64, 1.83) (1.63, 1.82) a: bootstrap with a confined parameter bound,
b: bootstrap with a large parameter bound, but bootstrap
estimates with parameter values hitting the right boundary are removed.
Mwd =2Nref×mwd, mwd is the fraction of the wolf population that is migrants from dogs. Nref is the population size for the reference population. Mdw is defined similarly.
Here, we set the reference population to be the ancestral population. Confidence interval is calculated as
the )(96.1 **
Supplementary Table S6: Regions show strong signal of positive selection
Chrom Start End Chrom Start End
chr1 4920000 5020000 chr12 34240000 34380000
chr1 5540000 5640000 chr12 38760000 38900000
chr1 5820000 6100000 chr12 41800000 41960000
chr1 6060000 6200000 chr12 44020000 44160000
chr1 6360000 6500000 chr12 44200000 44320000
chr1 6580000 6760000 chr12 50920000 51180000
chr1 6740000 6920000 chr12 51200000 51300000
chr1 24200000 24340000 chr12 52020000 52260000
chr1 28480000 28600000 chr12 58920000 59040000
chr1 35140000 35240000 chr13 11240000 11340000
chr1 58440000 58580000 chr13 53620000 53760000
chr1 61160000 61260000 chr14 12400000 12540000
chr1 66700000 66800000 chr14 12480000 12580000
chr1 68440000 68580000 chr14 58220000 58320000
chr1 82620000 82880000 chr14 58560000 58680000
chr1 82920000 83160000 chr15 8120000 8400000
chr1 83240000 83400000 chr15 8440000 8580000
chr1 86840000 86940000 chr15 20960000 21140000
chr1 87140000 87260000 chr15 38360000 38500000
chr1 94840000 94980000 chr15 40020000 40180000
chr1 102080000 102280000 chr16 8900000 9040000
chr1 114160000 114280000 chr16 9740000 10040000
chr2 6700000 6800000 chr16 10100000 10200000
chr2 6740000 6880000 chr16 17020000 17120000
chr2 10140000 10320000 chr16 19680000 19860000
chr2 10260000 10380000 chr16 29100000 29420000
chr2 39520000 39640000 chr16 44080000 44200000
chr2 52500000 52620000 chr16 44140000 44280000
chr2 52580000 52680000 chr17 8580000 8680000
chr2 55940000 56080000 chr17 24920000 25020000
chr2 56200000 56300000 chr18 3660000 3840000
chr2 60160000 60260000 chr18 3840000 4060000
chr3 20160000 20280000 chr18 4420000 4600000
chr3 21160000 21260000 chr18 4580000 4720000
chr3 21520000 21660000 chr18 5700000 5860000
chr3 43580000 43680000 chr18 5800000 5960000
chr3 54400000 54580000 chr18 6140000 6280000
chr3 59900000 60040000 chr18 6440000 6720000
chr5 5260000 5360000 chr18 6700000 6820000
chr5 6840000 7260000 chr18 6980000 7200000
chr5 9400000 9500000 chr18 7160000 7420000
chr5 21080000 21200000 chr18 15240000 15360000
chr5 21220000 21400000 chr18 18520000 18660000
chr5 21440000 21540000 chr18 27240000 27480000
chr5 38880000 38980000 chr18 28980000 29080000
chr5 43780000 43900000 chr18 45500000 45600000
chr5 44740000 44900000 chr18 46420000 46580000
chr5 45740000 45900000 chr18 56500000 56600000
chr5 67180000 67280000 chr19 6520000 6620000
chr5 67220000 67340000 chr19 6760000 6920000
chr5 67420000 67540000 chr19 14100000 14220000
chr5 80120000 80280000 chr20 3600000 3700000
chr5 80900000 81080000 chr20 3800000 3980000
chr6 15040000 15160000 chr20 27020000 27160000
chr6 15100000 15200000 chr20 44440000 44540000
chr6 31520000 31700000 chr21 4280000 4580000
chr6 32380000 32540000 chr21 6100000 6300000
chr6 40640000 40780000 chr21 9080000 9180000
chr6 50200000 50420000 chr21 10160000 10260000
chr6 52560000 52680000 chr21 38740000 38840000
chr7 16400000 16520000 chr22 17580000 17680000
chr7 19980000 20100000 chr22 20720000 20820000
chr7 39540000 39680000 chr22 31160000 31260000
chr7 46940000 47080000 chr22 33080000 33180000
chr7 56040000 56160000 chr22 38160000 38340000
chr7 56300000 56400000 chr22 44960000 45060000
chr8 16060000 16220000 chr22 47180000 47300000
chr8 17640000 17880000 chr24 11460000 11560000
chr8 24240000 24340000 chr24 11540000 11640000
chr8 37280000 37500000 chr25 3620000 3800000
chr8 37460000 37560000 chr25 8800000 8920000
chr8 39800000 39920000 chr25 16160000 16300000
chr8 53960000 54060000 chr25 16760000 16860000
chr8 65440000 65600000 chr25 19700000 19840000
chr9 29060000 29240000 chr25 21100000 21240000
chr9 43820000 43920000 chr26 5380000 5540000
chr9 47480000 47580000 chr27 8360000 8480000
chr9 58340000 58460000 chr28 9520000 9620000
chr10 6420000 6600000 chr28 17920000 18060000
chr10 6680000 7040000 chr30 8020000 8160000
chr10 9060000 9160000 chr30 8980000 9100000
chr10 14800000 15120000 chr30 9040000 9160000
chr10 16580000 16680000 chr31 4660000 4780000
chr10 18700000 18820000 chr31 4840000 4960000
chr10 18840000 18980000 chr31 6140000 6340000
chr10 47340000 47440000 chr31 9340000 9440000
chr10 47420000 47540000 chr31 20380000 20520000
chr10 47480000 47620000 chr32 3520000 3720000
chr10 47560000 47700000 chr32 27380000 27500000
chr10 47740000 47960000 chr33 5200000 5320000
chr10 49220000 49480000 chr34 17620000 17780000
chr11 3020000 3140000 chr36 5140000 5300000
chr11 5820000 6000000 chr36 6620000 6800000
chr11 6340000 6440000 chr37 3160000 3260000
chr11 14800000 14960000 chr37 7380000 7500000
chr11 22000000 22100000 chr37 8780000 8880000
chr11 49780000 49940000 chr37 10100000 10240000
chr11 54640000 54740000 chr37 13560000 13660000
chr11 56820000 57080000
chr11 57020000 57120000
Supplementary Table S7. GO analysis of positively selected genes
Go_term Pvalue
Fold
Enrichment
Biological Process
GO:0048609~reproductive process in a multicellular
organism 0.002 2.68
GO:0032504~multicellular organism reproduction 0.002 2.68
GO:0044058~regulation of digestive system process 0.006 25.44
GO:0007276~gamete generation 0.010 2.60
GO:0019953~sexual reproduction 0.010 2.44
GO:0009057~macromolecule catabolic process 0.019 1.91
GO:0008037~cell recognition 0.021 6.79
GO:0060457~negative regulation of digestive system
process 0.021 93.30
GO:0010949~negative regulation of intestinal phytosterol
absorption 0.021 93.30
GO:0045796~negative regulation of intestinal cholesterol
absorption 0.021 93.30
GO:0044265~cellular macromolecule catabolic process 0.022 1.93
GO:0016337~cell-cell adhesion 0.028 2.70
GO:0009063~cellular amino acid catabolic process 0.036 5.49
GO:0032368~regulation of lipid transport 0.040 9.33
GO:0033631~cell-cell adhesion mediated by integrin 0.042 46.65
GO:0009310~amine catabolic process 0.050 4.78
GO:0046777~protein amino acid autophosphorylation 0.062 4.39
GO:0030300~regulation of intestinal cholesterol
absorption 0.062 31.10
GO:0033627~cell adhesion mediated by integrin 0.072 26.66
GO:0032372~negative regulation of sterol transport 0.072 26.66
GO:0007158~neuron adhesion 0.072 26.66
GO:0032375~negative regulation of cholesterol transport 0.072 26.66
GO:0010604~positive regulation of macromolecule
metabolic process 0.072 1.63
GO:0007586~digestion 0.073 4.10
GO:0007368~determination of left/right symmetry 0.073 6.66
GO:0009799~determination of symmetry 0.077 6.51
GO:0009855~determination of bilateral symmetry 0.077 6.51
GO:0030299~intestinal cholesterol absorption 0.082 23.32
GO:0006281~DNA repair 0.083 2.30
GO:0006259~DNA metabolic process 0.091 1.84
GO:0009077~histidine family amino acid catabolic process 0.092 20.73
GO:0006548~histidine catabolic process 0.092 20.73
GO:0048610~reproductive cellular process 0.095 2.88
GO:0006508~proteolysis 0.095 1.50
Cellular Component
GO:0042995~cell projection 5.14E-04 2.43
GO:0031981~nuclear lumen 0.010 1.63
GO:0030424~axon 0.011 3.73
GO:0070013~intracellular organelle lumen 0.014 1.52
GO:0043233~organelle lumen 0.019 1.49
GO:0043005~neuron projection 0.019 2.48
GO:0031974~membrane-enclosed lumen 0.025 1.46
GO:0031594~neuromuscular junction 0.027 11.54
GO:0005654~nucleoplasm 0.054 1.63
GO:0044463~cell projection part 0.058 2.53
GO:0005929~cilium 0.065 3.28
GO:0005730~nucleolus 0.067 1.70
GO:0005911~cell-cell junction 0.073 2.67
Molecular Function
GO:0016894~endonuclease activity, active with either
ribo- or deoxyribonucleic acids and producing 3'-
phosphomonoesters 1.94E-07 26.15
GO:0004522~pancreatic ribonuclease activity 5.02E-07 35.03
GO:0016892~endoribonuclease activity, producing 3'-
phosphomonoesters 1.30E-06 29.50
GO:0004540~ribonuclease activity 6.08E-06 11.32
GO:0004518~nuclease activity 4.69E-05 5.91
GO:0004519~endonuclease activity 9.26E-05 7.47
GO:0004521~endoribonuclease activity 1.35E-04 11.92
GO:0005518~collagen binding 0.056 7.78
GO:0001640~adenylate cyclase inhibiting metabotropic
glutamate receptor activity 0.062 31.13
GO:0070742~C2H2 zinc finger domain binding 0.062 31.13
GO:0046982~protein heterodimerization activity 0.071 2.69
GO:0016888~endodeoxyribonuclease activity, producing
5'-phosphomonoesters 0.082 23.35
GO:0003684~damaged DNA binding 0.099 5.60
(cutoff pvalue at 0.1 level)
Supplementary Table S8. Genes found in human genome scans for positive selection
Gene NS Description Reference
ABCG5 4 ATP-binding cassette, sub-family G (WHITE), member
5
58-60,*
ABCG8 4 ATP-binding cassette, sub-family G (WHITE), member
8
58-60,*
ALS2CR11 2 amyotrophic lateral sclerosis 2 (juvenile) chromosome
region, candidate 11
61,*
BFAR 3 bifunctional apoptosis regulator 62-64
BRE 2 brain and reproductive organ-expressed (TNFRSF1A
modulator)
60,*
C11orf49 2 chromosome 11 open reading frame 49 60
,*
CENPP 2 centromere protein P 60
,*
CPEB3 2 cytoplasmic polyadenylation element binding protein 3 59,60
DYNC2LI1 4 dynein, cytoplasmic 2, light intermediate chain 1 58-60
,*
GRM8 2 glutamate receptor, metabotropic 8 60,61
HECW2 5 HECT, C2 and WW domain containing E3 ubiquitin
protein ligase 2
58-61,64
ITGB1 2 integrin, beta 1 59,60
LRPPRC 4 leucine-rich PPR-motif containing 58-60
,*
MET 2 met proto-oncogene (hepatocyte growth factor receptor) 60,61
MIR423 4 microRNA 423 58-60
,*
MOXD1 2 monooxygenase, DBH-like 1 58,61
MRPL46 3 mitochondrial ribosomal protein L46 61,63,65
MRPS11 3 mitochondrial ribosomal protein S11 61,63,65
NSRP1 4 nuclear speckle splicing regulatory protein 1 58-60
,*
PARN 3 poly(A)-specific ribonuclease 62-64
PLA2G10 3 phospholipase A2, group X 62-64
PLEKHH2 4 pleckstrin homology domain containing, family H (with
MyTH4 domain) member 2
58-60,*
PRSS1 6 protease, serine, 1 (trypsin 1) 58,61-64
,*
RBKS 2 ribokinase 60
,*
SLC6A4 4 solute carrier family 6 (neurotransmitter transporter,
serotonin), member 4
58-60,*
SPDYE2 4 speedy homolog E2 (Xenopus laevis) 60,61,65
,*
SPDYE6 4 speedy homolog E6 (Xenopus laevis) 60,61,65
,*
STK17B 5 serine/threonine kinase 17b 58-61,64
WDR75 2 WD repeat domain 75 58,60
ZMYM2 6 zinc finger, MYM-type 2 60-63,65
,*
ZMYM5 6 zinc finger, MYM-type 5 60-63,65
,*
ZNF786 2 zinc finger protein 786 61
,* *, Fst based calculation (unpublished) in Akey 2009 Genome Research
Supplementary Note 1. Sanger verification of Single Nucleotide Variants (SNVs)
Our genetic information (e.g. SNP calling) was extracted aggregating across all 11
individuals including the boxer reference. DNA for Sanger verification from all
individuals, however, is not available. Here, we performed the experimental verification
in a small dataset and extracted the trend for the larger sample. Thus, PCR primers were
designed to sequence genome segments across the dog genome for six individuals during
our whole genome sequencing. False positives and false negatives in SNP calling within
a single individual were examined first, followed by across all the 11 individuals.
After quality control, we were able to sequence, using Sanger methods, 614 segments
with a total length of 263,763bp (across all six individuals). For most amplicons we were
able to sequence all six individuals (Supplementary Table S2).
The levels of false positive and false negative SNP calling within each individual is
shown in Supplementary Table S3. False positives are quite low and false negatives are
around 0.2-0.3. A large proportion of the false negatives are due to low coverage in the
genome sequencing. If we restrict our analysis to parts of the genome that are sequenced
at higher coverage, the false negatives are much fewer.
Since errors are mostly random and non-overlapping, when multiple individuals are
sequenced, the false positives will increase, but false negatives decrease as germline
SNPs will be shared between individuals. Therefore, we randomly sampled 2-6 of the
Sanger sequenced individuals and examined the trend of false positives and false
negatives by changing the number of sampled individuals (Supplementary Figure S2).
As shown, the observed trends match our expectations. It should be emphasized that the
decrease in false negative is rapid while the increase in false positive is quite slow. If we
fit a line through the observed points, the predicted false negative should be less than
10%, while false positive should be no larger than 5% across 11 individuals
(Supplementary Figure S2).
A set of similar calculations was done for Indels. The calculated false positive is 6.5%
while false negative is 55% (Supplementary Table S4). The high false negative is mostly
due to the fact that we used very stringent criteria and only picked high quality indel
mutations.
Supplementary Note 2. PCA analysis
a) PCA analysis over all the canids
Genotype data was obtained from previous studies66,67
for 912 dogs, 209 grey wolves, 58
coyotes and 12 red wolves and combined with data obtained from the genomes reported
here for 11 individuals and one additional outgroup species (a red wolf that we sequenced)
for a total of 1203 individuals surveyed across the 48k SNP markers.
SmartPCA was employed to perform the PCA analysis across the 1203 individuals. From
Fig.2d we see that the split between the domesticated dogs and wolves is clear. The first
principle component, which accounts for 10.4% of the total variation, directly separates
these two groups. Interestingly, there is a group of dogs (we call this group 1) that are
closer to the wolves.
Our Chinese indigenous dogs are located within group 1. In addition to the Chinese
indigenous dogs, other breeds that are known to originate in China/Southeast Asia (e.g.,
Tibetan Mastiff (that we sequenced here), Chow-chow, Chinese Shar-Pei) are in group 1.
A few other breeds, for example, Dingo and New Guinea Singing Dog, are suggested to
have South-east Asian origin68,69
, while a few other breeds from Siberia (Siberian Husky
and Alaskan Malamute) and Japan (Akita) might also have a Southeast Asian origin70
.
All of these breeds are in group 1. The Basenji, an African dog that is often classified as
an ancient breed, is the only other breed found in group 1, and a previous mtDNA study
also suggested a closer relationship of this breed to many Chinese breeds71
. The exact
history of the Basenji, especially whether it has an Asian root is not very clear at this
moment.
Previous studies have argued for a Middle-Eastern origin of dog based on geographic
patterns from wolves, especially the fact that Middle Eastern wolves, as a group, seem to
be closer to dogs than wolves from other places using the 48K SNP chip data66,67
. There
are several potential confounding factors that might affect conclusions drawn from that
study.
1) Wolf populations have been greatly affected by human activities in recent history. For
example, the ancestral Chinese wolves for the domesticated dog may be extinct. It is
difficult to use patterns from extant wolves to infer the domestication location of an
ancestral population.
2) Several European wolves are found to be even closer to dogs, in spite of the fact that
European wolves as a whole are slightly further away from dogs than Middle Eastern
wolves (Fig.2d).
3) The Southwest Asian wolves are much further away from dogs, and are quite distinct
from Middle Eastern wolves, even though geographically they are very close (Fig.2d).
This suggests that there might be possible confounding factors contributing to the
discrepancy between genetic relationship and geographic location.
4) SNP ascertainment biases seem to be quite strong in the 48K SNP chip data. For
example, when we look at allele frequencies at the SNP positions in our resequenced data,
these SNPs are strongly enriched towards high frequency (Supplementary Figure S5). In
addition, SNP ascertainment biases are found to be correlated with the fact that these
SNP markers were developed during the sequencing of the first dog genome (a boxer
individual). In the PCA plot, the boxer/bull dogs group is distributed at some distance
from the other dog breeds (Supplementary Figure S6). SNP ascertainment biases might
have significantly contributed to PC2.
These observations indicate that demographic relationship of wolves is not yet resolved.
The argument for a Middle Eastern origin using patterns from wolves might be
confounded by these factors. These observations suggest that a careful examine of these
factors is needed before we can better explore the question of domestication location
using patterns from wolves.
b) Chinese indigenous dogs and native dogs from other geographic regions
Using mtDNA and Y chromosome data, several of the previous studies found that genetic
diversity is highest in Southeast Asia/China and is generally higher than populations from
the rest of the world 71-74
. One study that challenged this view is from Boyko et al 2009,
where the authors found that African village dogs have comparable genetic diversity as
those from Southeast Asia75
. However, this conclusion was later contested and was found
to be questionable73,74
. A recent study with many native dogs across much of the old
world also revealed a pattern with the highest genetic diversity in Southeast Asia 72
. In
addition, Chinese indigenous dogs together with several dog breeds originated from
Southeast Asia (often designated as ancient breeds) are found to be the most basal
lineages linking to grey wolves 71,73,74
.
Besides studies using only a few genetic markers, majority of the whole genome studies
are based on SNP chips. For example, several work from Wayne and his colleagues have
surveyed the genetic information among a global collection of more than 1000 canids 66,67,76
. In these studies, several dog breeds from Asia are found to be the first tier of
groups that are closest to wolves 67
. Majority of the dogs surveyed were breed dogs
except a few cases where the African village dogs were included66,67,76
.
When we extracted the genetic information, including all the dog breeds and the African
dogs, we found that, the African village dogs are much further away from the wolves
than the first tier of dogs, which include the Chinese native dogs sequenced here and the
ancient breeds (Supplementary Figure S7). It supports that Chinese indigenous dogs
together with a group of breed dogs that originated in Southeast Asia are the first tier of
dogs that are closest to wolves.
(*data were extracted from http://genomemirror.bscb.cornell.edu/cgi-bin/hgGateway)
Supplementary Note 3. Demographic analysis
When performing demographic analysis using dadi, there are various options for the
search space for the underlying parameters. In our analysis, we set the upper bound of t
to be 0.3 (equivalence of ~100,000 years), the maximum likelihood estimates show a
single sharp peak around the 0.1 (32,000 years). This parameter setting is equivalent for
using a hard bound (a fixed interval for the possible divergence time) similar to what is
typically conducted in molecular dating.
In the exploratory analysis, we noticed a phenomenon with the statistical model that is
not discussed in the literature very extensively and worth a discussion here. The
combination of parameters, in particular, migration rate (m) and divergence time (t) can
have multiple peaks when inspecting within a large parameter domain (e.g. when setting
the parameter boundary to be much wider). In particular, the likelihood function might
take a form f(x,y)=f(x+y, x*y), where x and y are two parameters. This form means that
the x and y are somewhat “equivalent”, in other words, a situation with a large x and a
small y might be as “good” (in the sense of the likelihood) as those with a small x and a
large y. The parameter t (divergence time) and m (migration rate) here seem to have a
form similar to this effect. Thus, there are multiple peaks in the likelihood surface with
different combinations of parameters.
Since this is purely a mathematical property, in practice, it can be treated in two ways.
One is putting a prior and modeling things in the Bayesian framework (e.g. similar to the
soft bound in the molecular dating literature). The other is to restrict the parameter bound
within biologically reasonable ranges (e.g. hard bound). In our setting, we know that very
deep divergence is not possible for domesticated species (e.g. through fossil and
archeological records).
In our analysis, when we set the upper bound of t to be 0.3 (equivalence of ~100,000
years). The maximum likelihood estimates show a single sharp peak around the 0.1
(32,000 years). When we expanded the upper bound of t to be 0.5 or 0.8 (equivalence of
160,000 or 250,000 years), the second parameter peak “appears” (Supplementary Figure
S8). For example, in the 100 bootstrap replicates, a significant proportion of the replicates
have point estimate of t to be 0.5 (the right boundary, the mean of the right peak
estimates is 0.499392).
The reality is that, in the parameter domain, there is a second peak at large t and the
estimated value is hitting the parameter boundary. When we set the upper bound to 0.8,
the observation is very similar. So, there are two peaks in the likelihood surface. One of
the peaks for t is at 0.1, the other is at a very large t. Since the maximum likelihood
optimizer (search function) is doing a local hill climbing. Which peak it finds (absorbing
to) critically depends on the starting value. We found that, for a given bootstrap sample,
if we run the optimizing a lot more times (corresponding to different starting values), the
number of bootstrap samples that “absorb” to the right boundary will also decrease.
Through our limited explorations, we find that the peak might be much larger than 0.8.
(*The estimates keep hitting the right bound when we extend the upper bound. However,
the program becomes very slow when the bound is set to be very large, say >2 due to the
large search space). Since we know that very deep divergence is not possible for these
domesticated species, we thus removed the estimates from these bootstrap replicates
(Equivalence of implementing a hard bound). The maximum likelihood estimates for the
rest of the bootstrap samples are very similar to the case when we set the upper bound to
be 0.3 (Supplementary Table S5).
When we extracted site frequency spectra from the sequenced individuals, we restricted
ourselves to the non-coding part of the genome that is sequenced to a higher coverage
(See maintext). We are interested in quantify possible ascertainment biases in our
extracted site frequencies in addition to Sanger sequencing verifications.
We employed methods developed in a previous study, which uses a maximum likelihood
method to estimate allele frequencies in low and medium coverage next-generation
sequencing data. This method is based on integrating over uncertainties in the data for
each individual77
. When we compared our extracted Site Frequency Spectra (SFS) with
the inferred SFS using this maximum likelihood method, we found that our results match
quite well with the inferred SFS as well as the best fitted model (Supplementary Figure
S9 and S10).
Supplementary Note 4. Diversity estimates and genome wide plots Since the dogs in our sample are quite diverse, and do not come from the same
population, the heterogeneity of the individuals will prevent sophisticated population
genetic methods from identifying traces of adaption. We thus looked for the footprints of
adaption by looking for: 1) focal regions that show reduced genetic diversity in the dog
population (we used the bottom 5% quantile from the dog mean genome wide
distribution), 2) segments are not in a low diversity regions in wolves (we used the
bottom 20% quantile from the wolf mean genome wide distribution), 3) there is a high
divergence between the dog and wolf populations (we used the top 95% quantile in the
Fst distribution as the cutoff).
A full display of the diversity and cutoffs are presented in Supplementary Figure S11 and
the associated genomic regions are listed in the Supplementary Table S6.
Supplementary Note 5. Genes of interest in addition to those in the main text
Several genes (Supplementary Table S7), in particular genes for metabolic enzymes are
in our list of overlapping genes. For example, PLA2G10 is an important phospholipase
involved in the biochemical pathways transforming phospholipids and other lipophilic
molecules78
. PRSS1 gene is a trypsinogen gene79
and RBKS is a ribokinase80
. All of these
proteins play very important and diverse biological functions in processes such as
inflammation, cell growth, signaling and death and maintenance of membrane
phospholipids.
In addition to genes such as MET, other cancer-related genes involved in the apoptosis
pathway were found. For example, ITGB1 is a member of integrin family involved in the
metastatic diffusion of tumor cells 81
. BFAR, a gene that have been found in five human
genome scans, is an important apoptosis regulator82
. The BRE gene is part of a protein
complex that is involved in DNA damage and may also act as a death receptor-associated
anti-apoptotic protein83
. STK17B is a positive regulator of apoptosis84
. Natural selection
on apoptosis related genes on the human lineage was noticed in previous studies, e.g. 85
.
ZMYM2 is a transcription factor and translocation of this gene with fibroblast growth
factor receptor-1 gene (FGFR1) can result in a fusion gene that is associated with the
stem cell leukemia lymphoma syndrome 86
.
Supplementary References
58 Tang, K., Thornton, K. R. & Stoneking, M. A New Approach for Using Genome
Scans to Detect Recent Positive Selection in the Human Genome. PLoS biology 5,
e171 (2007).
59 Voight, B. F., Kudaravalli, S., Wen, X. & Pritchard, J. K. A map of recent
positive selection in the human genome. PLoS biology 4, e72 (2006).
60 Wang, E. T., Kodama, G., Baldi, P. & Moyzis, R. K. Global landscape of recent
inferred Darwinian selection for Homo sapiens. Proc Natl Acad Sci U S A 103,
135-140 (2006).
61 Kimura, R., Fujimoto, A., Tokunaga, K. & Ohashi, J. A practical genome scan for
population-specific strong selective sweeps that have reached fixation. PLoS ONE
2, e286 (2007).
62 Carlson, C. S. et al. Genomic regions exhibiting positive selection identified from
dense genotype data. Genome research 15, 1553-1565 (2005).
63 Kelley, J. L., Madeoy, J., Calhoun, J. C., Swanson, W. & Akey, J. M. Genomic
signatures of positive selection in humans and the limits of outlier approaches.
Genome research 16, 980-989 (2006).
64 Sabeti, P. C. et al. Genome-wide detection and characterization of positive
selection in human populations. Nature 449, 913-918 (2007).
65 Williamson, S. H. et al. Localizing recent adaptive evolution in the human
genome. PLoS genetics 3, e90 (2007).
66 vonHoldt, B. M. et al. A genome-wide perspective on the evolutionary history of
enigmatic wolf-like canids. Genome research 21, 1294-1305 (2011).
67 Vonholdt, B. M. et al. Genome-wide SNP and haplotype analyses reveal a rich
history underlying dog domestication. Nature 464, 898-902 (2010).
68 Oskarsson, M. C. et al. Mitochondrial DNA data indicate an introduction through
Mainland Southeast Asia for Australian dingoes and Polynesian domestic dogs.
Proc Biol Sci 279, 967-974 (2012).
69 Savolainen, P., Leitner, T., Wilton, A. N., Matisoo-Smith, E. & Lundeberg, J. A
detailed picture of the origin of the Australian dingo, obtained from the study of
mitochondrial DNA. Proc Natl Acad Sci U S A 101, 12387-12390 (2004).
70 Parker, H. G. et al. Genetic structure of the purebred domestic dog. Science (New
York, N.Y 304, 1160-1164 (2004).
71 Savolainen, P., Zhang, Y. P., Luo, J., Lundeberg, J. & Leitner, T. Genetic
evidence for an East Asian origin of domestic dogs. Science 298, 1610-1613
(2002).
72 Brown, S. K. et al. Phylogenetic distinctiveness of Middle Eastern and Southeast
Asian village dog Y chromosomes illuminates dog origins. PLoS One 6, e28496
(2011).
73 Ding, Z. L. et al. Origins of domestic dog in southern East Asia is supported by
analysis of Y-chromosome DNA. Heredity 108, 507-514 (2012).
74 Pang, J. F. et al. mtDNA data indicate a single origin for dogs south of Yangtze
River, less than 16,300 years ago, from numerous wolves. Molecular biology and
evolution 26, 2849-2864 (2009).
75 Boyko, A. R. et al. Complex population structure in African village dogs and its
implications for inferring dog domestication history. Proceedings of the National
Academy of Sciences of the United States of America 106, 13903-13908 (2009).
76 Boyko, A. R. et al. A simple genetic architecture underlies morphological
variation in dogs. PLoS biology 8, e1000451 (2010).
77 Kim, S. Y. et al. Estimation of allele frequency and association mapping using
next-generation sequencing data. BMC Bioinformatics 12, 231 (2011).
78 Cupillard, L., Koumanov, K., Mattei, M. G., Lazdunski, M. & Lambeau, G.
Cloning, chromosomal mapping, and expression of a novel human secretory
phospholipase A2. J Biol Chem 272, 15745-15752 (1997).
79 Emi, M. et al. Cloning, characterization and nucleotide sequences of two cDNAs
encoding human pancreatic trypsinogens. Gene 41, 305-310 (1986).
80 Park, J., van Koeverden, P., Singh, B. & Gupta, R. S. Identification and
characterization of human ribokinase and comparison of its properties with E. coli
ribokinase and human adenosine kinase. FEBS Lett 581, 3211-3216 (2007).
81 Garmy-Susini, B. et al. Integrin alpha4beta1 signaling is required for
lymphangiogenesis and tumor metastasis. Cancer Res 70, 3042-3051 (2010).
82 Prat, M. et al. The receptor encoded by the human c-MET oncogene is expressed
in hepatocytes, epithelial cells and solid tumors. Int J Cancer 49, 323-328 (1991).
83 Li, Q. et al. A death receptor-associated anti-apoptotic protein, BRE, inhibits
mitochondrial apoptotic pathway. J Biol Chem 279, 52106-52116 (2004).
84 Sanjo, H., Kawai, T. & Akira, S. DRAKs, novel serine/threonine kinases related
to death-associated protein kinase that trigger apoptosis. J Biol Chem 273, 29066-
29071 (1998).
85 da Fonseca, R. R., Kosiol, C., Vinar, T., Siepel, A. & Nielsen, R. Positive
selection on apoptosis related genes. FEBS Lett 584, 469-476 (2010).
86 Lierman, E. & Cools, J. Recent breakthroughs in the understanding and
management of chronic eosinophilic leukemia. Expert Rev Anticancer Ther 9,
1295-1304 (2009).